Spectral calculus and Lipschitz extension for barycentric metric spaces (Q2851032)

The main purpose of this very interesting paper is to compute a bi-Lipschitz invariant, called metric Markov cotype, for barycentric metric spaces. From this result the authors deduce several important corollaries. We refer the reader to the paper for the precise formulations of notions and results mentioned below.NEWLINENEWLINEFor a metric space, the notion of Markov type \(p\) introduced by \textit{K. Ball} [Geom. Funct. Anal. 2, No. 2, 137--172 (1992; Zbl 0788.46050)] is recalled. A definition of metric Markov cotype \(p\) is then given and it is explained why this notion is equivalent with the definition of Markov cotype \(p\) introduced in the above paper of K. Ball.NEWLINENEWLINEThe authors introduce a notion of a barycenter map and define \(W_p\) barycentric and \(p\)-barycentric metric spaces. The class of spaces, which are \(W_p\) barycentric and \(q\)-barycentric with respect to the same barycenter mapping, includes all complete \(CAT(0)\) metric spaces (Hadamard spaces). Banach spaces whose modulus of convexity have power type \(p\in [2,\infty)\) are \(W_1\) barycentric and also \(p\)-barycentric.NEWLINENEWLINEThe first result of the paper states that a metric space, that is \(W_p\) barycentric and also \(p\)-barycentric, has metric Markov cotype \(p\). As the first application of this theorem the paper presents results on calculus for nonlinear spectral gaps. The second application is to the problem of Lipschitz extension; the authors proves a nonlinear version of Ball's extension theorem. This result provides a versatile Lipschitz extension theorem that encompasses a wide range of seemingly disparate Lipschitz extension results.NEWLINENEWLINENext they use \textit{N. J. Kalton}'s work in [Math. Ann. 354, No. 4, 1247--1288 (2012; Zbl 1268.46018)], to construct a closed linear subspace of \(\ell_1\) that fails to have metric Markov cotype \(p\) for every \(p\in (0,\infty)\). This shows that for Banach spaces metric Markov cotype and Rademacher cotype are different notions. An interesting byproduct of the quantitative analysis of Kalton's construction is that it shows that a Lipschitz extension result by Johnson, Lindenstrauss and Benyamini (see [\textit{W. B. Johnson} and \textit{J. Lindenstrauss}, Contemp. Math. 26, 189--206 (1984; Zbl 0539.46017)] and [\textit{W. B. Johnson} et al., Isr. J. Math. 54, 129--138 (1986; Zbl 0626.46007)]) is asymptotically sharp.